Correct the water backflow into cavern to compensate for unmodelled creep

[dbhart]

Summary
When injection stops and the cavern is "shut in", the code compensates for pressure losses from dissolution by flowing water backwards into the system. While this maintains a mass balance, this is physically incorrect, as in reality there is decreasing pressure within the cavern that makes up the difference instead; sansmic has never accounted for pressure.

This seems to be less of a problem in ordinary leach modes and the greatest problem in withdrawal and leach-fill modes.

Possible solution
One possible solution would be to track the volume pulled back into the system, ensure that it does not get flagged as injected water (which has been creating non-physical, one or two cell flares at the OBI), but make the information about that volume available. If desired, someone could create a very simple creep model plugin that could convert that into an "expected pressure change" output.